</> MAANG.io

Foundations

The behavioral questions new grads and SDE I candidates are asked, from tell me about yourself to your first mistakes and disagreements. Every lesson also shows how the expected answer grows from SDE to Principal.

0 of 36 solved 0%

Tell Me About a Time You Had to Work on Multiple Projects at the Same Time

The question: "Tell me about a time when you had to work on multiple projects or tasks at the same time. How did you decide what to work on first?"

Also asked as: "How do you decide what to work on when everything is urgent?" · "Tell me about a time you had to drop something. Who did you tell?" · "How do you choose what not to do?"

Who's asking: an Engineering Director or Senior Engineering Manager, in a behavioral round.

Executive Summary: The director isn't checking that you have a framework. They're checking that you can say what you didn't do, who you told, and whether they heard it from you before it slipped. What changes most with level is whose list it is: your week, your team's quarter, several teams' asks on one shared team, the org's year.

What the director is really asking

Every director has had a report who says yes to everything, then goes quiet. So they're answering three questions:

  • "What fell off?" If everything got done, the list wasn't real or you're not telling them.
  • "Who heard it from you?" A slip is forgiven. A surprise isn't. They want the date you told the person waiting, next to the slip.
  • "What were you ranking against?" One question or one number, said out loud. A four-quadrant matrix is a method, not a decision.

From Staff up, there's a fourth: "Whose list was it?" At Staff the asks come from teams you don't run, and the choice has no owner. At Principal the list is what the org funds this year, on the customer's clock.

It maps most closely to Amazon's "Deliver Results" (approximate), including the half nobody quotes: saying what won't be delivered. Three ladder dimensions do most of the work: ownership (raising risks early, then owning outcomes beyond your authority), time horizon and ambiguity (handed the tasks, or finding the choice nobody owned).

Level fit at a glance

SDE Senior SDE Staff Principal
Asked at this level? Core Core Common Rare
Story scope they expect Three things on your own plate, inside your team's system, over a week or two Your team's quarter: items that each want the whole team, with a customer or partner team affected Several teams' asks on something shared, where nobody owned the order Which programs the org funds this year, under your name
The bar in one line Rank against one question with your lead, tell whoever's waiting before it slips, own what you got wrong Rank the quarter against stated costs, own the slip to whoever it hurts, watch the number you chose Find the ranking nobody owns, get managers you don't run to accept it, leave it with an owner Set the year's order with the business clock on it, say what it cost, leave a mechanism in the funding process
Instant down-level "I got all three done by working nights" "I asked my manager which came first" "I prioritized my team's roadmap" "I ranked the programs by dependency"

Core at SDE and Senior, where a quiet yes-sayer costs a team a quarter. At Staff it arrives as "how do you choose which cross-team problem to take". At Principal, in this wording, it's rare: the director asks what the org works on next.

SDE: rank against one question, and say what slips before it does

What meets the bar

  • You ranked against one question you can repeat, and took it to your lead.
  • You told the person waiting on day one, not the day it slipped.
  • You own the thing you got wrong, with the time it cost.
  • The one you put last slipped, and you can say what that cost someone else.
  • Check your altitude: doing all three by working late reads below the bar, because nothing was decided. Your team's quarter, with a commitment re-cut with a customer, reads as Senior.

Model answer

"I was about 18 months in, on the six-person team that routes orders from a florist delivery network's website to about 1,400 partner shops. The week before Mother's Day is about a third of our year. That Monday I had three things. I was secondary on-call, and a bug had come in: orders in some rural areas were being rejected by the shop we'd routed them to. I'd promised the PM a delivery-window picker for the holiday order page, due before Wednesday's code freeze. And I'd promised a runbook for our routing service to the two engineers covering holiday on-call, neither of whom had run it.

I asked my lead one question: which of these gets worse if it waits a week? The picker didn't. The bug did, with volume climbing daily, and the runbook did, because the people who'd need it were on shift Saturday. So the order was bug, runbook, picker, and I told the PM that morning that the picker would miss the freeze.

Then I got the bug wrong. The rejected orders were rural, so I decided our address geocoding was off and spent a day and a half building a re-check. It wasn't that. Shops shrink their delivery radius in peak week through the partner portal, and our routing cache held the old radius for 24 hours. I found it when I finally pulled ten rejected orders with partner support and looked at what each shop had changed that day. The fix was a two-hour cache change. I'd had the order right and the problem wrong, and I told my lead the day and a half was mine.

Rejections in those areas went from about 4 percent of orders to under 1 before the weekend. The holiday on-call paged once, on the Saturday, followed the runbook and had it closed in 20 minutes. The picker shipped two weeks after the holiday, and the PM had to pull it from a marketing email they'd already drafted. What I do differently: before I decide what a bug is, I pull ten examples and look at what changed in them. Last month a double-charge bug looked like a retry problem until the samples showed every one was a double-click."

Why this lands at SDE

  • "I asked my lead one question: which of these gets worse if it waits a week?": this earns the level. One question, taken to the lead, and an order that came out of it.
  • "I told the PM that morning that the picker would miss the freeze": the person waiting heard it on day one, from the candidate.
  • "I'd had the order right and the problem wrong, and I told my lead the day and a half was mine": the miss is theirs, sized and reported.

Follow-ups to expect

  • "You spent a day and a half on the wrong cause. When did your lead find out?": tests whether you raise risks early → the day you told them, and don't shrink the gap between the two dates.
  • "What did the PM say on Monday?": tests how you treat people when you're the bad news → what they said, and what they undid.
  • "If the bug had turned out small, would you still have dropped the picker?": tests whether the order was reasoned or reactive → yes or no, and the question you'd have asked again.

Red flags

  • ❌ "I stayed late and got all three done." Nothing was decided, and nobody believes it.
  • ❌ "The PM kept changing priorities." The PM is now the villain in a story about your judgment.

Senior SDE: rank the quarter, own the slip, and watch the number you chose

What meets the bar

  • You ranked your team's quarter against a stated cost per item, and you can say which one you put last and why.
  • You re-cut a commitment yourself, with the customer or partner team it hurt, rather than letting it slip.
  • You name the number you tracked and what it hid.
  • The thing you parked cost something, and you say what.
  • Check your altitude: three things on your own plate with a lead deciding reads as SDE. Asks from teams you don't run, on a shared team nobody ranked, read as Staff.

Model answer

"I led the team of eight that owns search for a legal-research product, about 40,000 lawyers and paralegals. One quarter, three things each wanted the whole team. The platform team was retiring the cluster version our index ran on at quarter end. An incident review had committed us to an alert on stale index refreshes, after search served three-hour-old results one afternoon. And sales had promised our largest firm, 1,200 seats up for renewal mid-quarter, search by case citation.

I wrote one line per item: what happens if it slips a month. The migration had a date nobody could move. The renewal had a dollar figure. The alert had neither, so I ranked it third, and that's the one that bit.

I tracked the migration by percent of documents moved. At week six it read 80, so I moved two engineers to citation search. The last 20 percent was the old statute corpus, on a different schema, and in week nine the engineer on it told me the remaining shards needed remapping. By effort we were about half done. I'd picked a number that hid the hard part. The same week the index went stale for 40 minutes mid-migration, and a customer's email found it, not us. Exactly the gap the parked alert would have closed.

I went to my manager and the PM with three moves. I asked the platform team for three weeks on the old cluster and paid for it from our budget, about $18,000. I called the firm's knowledge manager with the account manager and offered the two citation formats that covered about 85 percent of their lookups by renewal, the rest six weeks after. And I gave the alert to a junior engineer. It took three days.

The migration finished two weeks past quarter end. The firm renewed, about $900,000 a year, and the remaining formats shipped on the date I'd given. The alert has fired twice since and caught it inside five minutes. Now I track by effort left, not units moved, and I ask the engineer doing the work, not the dashboard. On a pipeline move this year I counted by table complexity, and the last 10 percent was budgeted at 40 percent of the time."

Why this lands at Senior SDE

  • "I called the firm's knowledge manager with the account manager and offered the two citation formats that covered about 85 percent of their lookups by renewal, the rest six weeks after": this earns the level. They re-cut the commitment themselves, with the customer, in the customer's terms.
  • "I'd picked a number that hid the hard part": the miss is in their own measurement, and they say so before anyone asks.
  • "Exactly the gap the parked alert would have closed": the thing they ranked last cost something, and the result is net of the $18,000.

Follow-ups to expect

  • "You parked the alert, and then the thing it would have caught happened. How did you explain that to your manager?": the hardest one here → tests whether you own the ranking, not just the fix. Three days of work you'd put behind two bigger things, and what you now do with the small cheap item.
  • "What did the firm's knowledge manager say to two formats?": tests how you treat the customer when you're the bad news → what they asked for in return.
  • "Who did you grow on this?": tests whether you built the team, not just the plan → the junior who built the alert: what you handed them (the incident review, the threshold choice) and what they own now (index freshness on-call).

Red flags

  • ❌ "I asked my manager which came first." That's an SDE move at Senior.
  • ❌ "Sales shouldn't have promised the feature." Now sales is the villain.

Quick check before moving on:

Staff: find the ranking nobody owns, and leave it with someone

What meets the bar

  • The asks came from teams you don't run, and nobody owned the order.
  • You ranked against one number you went and found, that no team's dashboard showed.
  • Managers you don't run accepted an order they didn't pick, and you can say what each gave up.
  • The order has an owner now, and it has since ranked something above your own ask.
  • Check your altitude: your own team's roadmap, however big, reads as Senior. If the choice moved what the business funds or how it earns, it reads as Principal.

Model answer

"A wine-and-spirits distributor, about 30,000 accounts. Four product teams built on one shared data team of six: catalog, pricing, state shipping rules. Each had an ask in its queue, ordered by whoever escalated last. Nobody owned the choice. I was the platform org's staff engineer, with no authority over any of them.

I handed my platform cost review to a senior engineer and spent the month with finance pulling a quarter of failed orders for one number to rank against. The biggest was stale state-shipping flags: 3.1 percent of self-service orders rejected at the warehouse; most re-keyed by a rep, one in five never came back, about $1.1 million a year in rep time and lost orders. One ask touched it, the self-service team's rules refresh. The loudest, the rep app's catalog search, didn't. I took it to all five managers; the queue was re-ordered.

The rep-app manager then went to the VP, and the queue reverted. I'd read their pushback as urgency. They were measured on the share of orders placed through the rep app, which the refresh would cut. I'd argued dollars at someone whose dollars weren't what was being counted.

I went back, asked what they reported up, and showed them the re-keying happened in their app, about two hours a day per rep they'd get back. I proposed to the VP, with both managers in the room, that the data team's manager own the queue, ranked by failed-order dollars each product manager supplied themselves. They agreed once catalog search was second, with a date.

Rejections went from 3.1 percent to 0.4 in a quarter, about $800,000 a year net of the two data-team engineers who shipped it in six weeks with self-service. Catalog search shipped a quarter late, and rep-app adoption dipped two points, which I own. Promotions missed quarter-end and ran their push on a spreadsheet. The ranking has run a year under the data team's manager, and last spring it put the warehouse team's ask ahead of one of mine. Now I ask each manager what number they report up before I rank across teams, and make the case in that number first. On a pricing-feed ask this year, that was audit findings, not speed."

Why this lands at Staff

  • "I proposed to the VP, with both managers in the room, that the data team's manager own the queue, ranked by failed-order dollars each product manager supplied themselves": the Staff line. A choice nobody owned gets an owner and a rule, and managers outside the candidate's authority accept it. No Senior story has this.
  • "I'd argued dollars at someone whose dollars weren't what was being counted": the miss is in how they read a person, and they went back rather than up.
  • "last spring it put the warehouse team's ask ahead of one of mine": the mechanism has an owner, and it has ruled against them.

Follow-ups to expect

  • "Why was this your problem? The data team has a manager.": the hardest one here → tests ownership beyond authority → the manager owned the queue, nobody owned the choice, and what you stopped doing to make room.
  • "The rep-app manager went over you to the VP. What did you say to them afterwards?": tests how you treat people → that you'd asked them for the wrong thing, and what you asked next.
  • "The promotions team ran quarter-end on a spreadsheet. Would you rank it the same way again?": tests judgment → yes, with the number, and what you'd tell them sooner.

Red flags

  • ❌ "I prioritized my team's roadmap for the year." That's a Senior story told at Staff.
  • ❌ "The rep-app manager was playing politics." A villain where an incentive was.

Principal: decide what the org works on next, with the customer's clock on it

What meets the bar

  • Why it's rare: a Principal isn't juggling tasks, and a director who asks this in this wording is checking whether the candidate still thinks in tasks. It gets asked instead as "what does the org work on next?" or "what did you choose not to fund?"
  • The list is the org's year, and the order went up under your name.
  • You ranked with the business clock, the dates customers and regulators decide, not only the dependency graph.
  • You say what the order cost in business terms, went back to the exec who funded it, and left a mechanism in the funding process.
  • Check your altitude: a VP in the room doesn't make it Principal. If the order was technical dependency alone and no funding moved, it reads as Staff.

Model answer

"I was the principal engineer for a national postal service's parcel-tracking platform, about 900 million parcels a year, 200 engineers. Five programs wanted funding that year: a mandated customs-data change, a scan backbone that missed about 2 percent of scans, delivery-time estimates the large shippers wanted, a shipper API and returns tracking. I set the order and took it to the COO under my name: customs first, backbone second, estimates third, because estimates built on missed scans are wrong.

That order was right for the system and wrong for the business. The backbone was 18 months. Two of our five largest shippers had contracts up at month nine, and a rival was offering estimates. I'd ranked by dependency and never put the date the customers decide next to it.

At month six, when the commercial director told me the rival had priced all five, I went back to the COO with a re-cut: a three-month estimate the estimates team built on the scans we had, right 78 percent of the time, with the backbone pushed four months. The two had signed with the rival before it shipped, about 15 percent of their volume, roughly $30 million a year. It held the other three, who'd had the same offer, about $80 million a year the rival was bidding for. The funding ranking now carries a column the commercial director owns: the date each customer decides. This year that column put the shipper API ahead of returns."

Why this lands at Principal

  • "That order was right for the system and wrong for the business": this earns the level. The miss is in the org's year, under their name, and they measure it in revenue, not velocity.
  • "I went back to the COO with a re-cut": they own the bet to the person who funded it, with an option they can price.
  • "a column the commercial director owns": the mechanism lives in the funding process, with an owner outside engineering.

Follow-ups to expect

  • "Two shippers left. What did the COO say, and would they take your ranking again?": the hardest one here → tests accountability for the bet → what they said, and what changed in how the order is built now.
  • "How did the backbone program's lead take four months of delay?": tests how you treat people when your correction lands on them → what you told them, and what they got.
  • "What mechanism outlived your involvement?": tests lasting change → the column, who owns it, and the one time it overruled you.

Red flags

  • ❌ "I ranked the programs by technical dependency." A tech lead's order, no business clock.
  • ❌ "The exec team chose the order." Then it wasn't your bet.

Quick check before moving on:

One story, four altitudes

SDE Senior SDE Staff Principal
Situation Bug, picker and runbook, the week before the holiday Migration, incident follow-up and a renewal, in one quarter Four teams' asks on one shared data team, ordered by escalation Five programs, 200 engineers, one year of funding
Your role SDE on the routing team, with a lead who can say no Leads the team of eight that owns search Staff engineer, no authority over any of the five teams Principal engineer, the year's order under your name
The key move Asked the lead "which gets worse if it waits?"; told the PM Monday Re-cut the renewal with the customer; paid for the extension Found the 3.1%, misread a manager, went back, got the queue an owner Set the order, missed the customers' clock, re-cut it with the COO
Result Rejections 4% to under 1%; a day and a half lost; picker two weeks late Firm renewed ($900k); $18k and two weeks; one 40-minute stale window 3.1% to 0.4%, about $800k a year net; adoption down two points for a quarter Held $80M of volume, lost $30M; backbone four months late
Reflection Pull ten examples before deciding what a bug is Track by effort left, and ask the engineer, not the dashboard Make the case in the other manager's number before dollars Put the customer's decision date in the ranking, next to dependency

Draft your answer

Step 1: Pick the right story

Run each candidate story through the test for your target level:

  • SDE: Were there at least three real things on your own plate, did you set the order with your lead, and did you tell whoever was waiting before their thing slipped?
  • Senior SDE: Was it your team's quarter, did you re-cut a commitment yourself with the customer or partner team it hurt, and did you own the launch and the slip?
  • Staff: Did the asks come from teams you don't run, on something shared where nobody owned the order, and did their managers accept an order they hadn't picked?
  • Principal: Was it which programs the org funds this year, under your name, with a business number on what the order cost?

Then check all of these:

  • It's from the last three years, or a longer Staff or Principal story you can bring up to date.
  • Something was put last and it slipped. If nothing slipped, pick another story; this one has no decision in it.
  • You can name the date you told the person waiting, and it's before the slip.
  • Something went wrong, and part of it was your doing: a wrong cause, a number that hid the problem, a misread person.

If you're a new grad. A research-lab job, a TA role or a part-time job counts if at least two people were waiting on you and one got their thing late; your own exams don't. A director accepts a shorter list, but still wants the rule you ranked by and the day you told the person you put last. "Two PhD students in my lab needed results the same week, and I had one GPU allocation."

Aiming a level above your current title? Don't inflate the scope. A Senior aiming at Staff can tell their team's quarter and show they've seen further: "I ranked my team's three, but the queue that mattered was the platform team's, and I'd now ask what their manager ranks against." That beats a story that falls apart at "Why was this your problem to solve?".

Step 2: Fill the skeleton

THE LIST (15s)       — What was on it, how big each item was, and whose list it
                       was (yours, your team's, several teams', the org's): ______
THE RULE (15s)       — The one question or number you ranked against, and why: ______
THE ORDER (15s)      — What went first, what went last, and what you told whoever
                       was waiting, and when: ______
WHAT WENT WRONG (25s) — The call that was wrong, what it cost, what you did next: ______
LEVEL BEAT (inside the above) — the line only your level could say.
  SDE: the question you took to your lead, and the slip you reported first.
  Senior: the commitment you re-cut yourself, with whoever it hurt.
  Staff: the order managers you don't run accepted, and who owns it now.
  Principal: the year's order under your name, and what it cost the business.
OUTCOME (25s)        — Net: what got done, what didn't, what the slip cost: ______
LEARNED (10s)        — What you do differently now, and one time you did: ______

Aim for about two minutes, with a 30-second cut ready (the list, the rule, the slip, the net result).

Step 3: Calibrate it

If your answer says… It reads as… Fix
A matrix or framework, with nothing named as slipping A method, not a decision The item you put last, its cost, and who you told
"Everything landed on time" Not believable The net result, including what got worse
A Staff answer: "I ranked my team's roadmap" Senior Whose asks competed, and what their managers accepted
A Principal answer: "I ranked the programs by dependency" Staff The customer's decision date, and the funding call under your name

Step 4: Rehearse the follow-ups

Prepare these five, and say each out loud once:

  1. "What fell off the list, and who heard it from you first?" The most dangerous one, at every level. Name the item, the person and the date, and don't soften the slip.
  2. "What were you ranking against?" One question or one number. If it was a gut call, say so, and what you'd rank against now.
  3. "The thing you put last, what did that cost?" A number or an incident, and whether it was worth it.
  4. "What did your manager know, and when?" The date you told them, and the gap to the slip. Don't shrink it.
  5. "What would you do differently?" A sharpening, with one time you've done it since. Not a confession.

Common mistakes

  • The quiet slip. "It ended up shipping later", nobody told. The director wants the date you said so.
  • A villain on the list. The PM who kept adding things, the rep who over-promised, the manager who escalated. Say what they were measured on instead.

Key takeaways

  • It tests whether you can say what you didn't do, who you told, and when, not whether you have a system.
  • As level rises, whose list it is grows: your week, your team's quarter, several teams' asks, the org's year.

Test yourself on this question's key ideas:

Sign in to MAANG.io

Use your Google or Microsoft account — no password to remember.

Continue with Google Continue with Microsoft

Please accept the terms above to continue.